Launch After Class: learning to test an idea without mistaking a simulation for a market

Prepared by Education Angel Group (EAG), with AI-assisted research and drafting. Research date: September 19, 2026. EAG is developing Launch After Class and has an interest in its success. This is a focused research report about a proposed learning experience, not an independent evaluation or evidence of product availability.

1. Proposition, audience and the proposed first experience

Launch After Class is a school-safe virtual-business game for practicing evidence-seeking and revision. The proposed first experience begins with one fictional customer problem. A student states a hypothesis, chooses a bounded test, examines modeled evidence and disclosed time or cost consequences, then revises or parks the hypothesis. The learning target is an explanation of what changed and why. Completing a virtual venture is not evidence that a real business would succeed.

The intended audiences have different jobs. A student needs a manageable decision and permission to change their mind. A teacher, parent or coach needs to understand the activity's reading demands, effort, privacy boundaries and educational purpose. A scenario author needs to make the model's assumptions inspectable. These requirements follow from EAG's current product definition, not from observed use of Launch.

The scope excludes real commerce, customer outreach, fundraising, investor simulation presented as reality, incorporation, payments and professional business guidance. The proposed problem, test choices, consequence rules and feedback are still design decisions. This report does not select an engine, approve a final scenario or establish an age range. It examines whether the narrow learning job has a defensible rationale and how to investigate it honestly.

Overview. Existing studies offer reasons to take hypothesis formation and revision seriously, but they do not establish the effectiveness of this children's game. Some entrepreneurship programs have changed selected outcomes; other measured outcomes have not improved. Established educational alternatives already exist. Launch's potential contribution is a small, repeatable practice opportunity whose evidence limits remain visible. Its central risk is that a persuasive simulation teaches students to treat supplied fictional responses as real demand.

2. Questions and standards of evidence

The governing question is: what helps students test and revise a customer/problem hypothesis without promoting unrealistic entrepreneurship claims? Five practical questions follow.

1. What is the learning need, and can it be separated from encouraging students to become entrepreneurs?
2. Which findings concern children's learning, which concern college education, and which concern decisions in adult firms?
3. What do established courses and an ordinary text exercise already offer?
4. How could modeled feedback make revision understandable while preserving uncertainty?
5. What evaluation would distinguish learning from completion, confidence or enthusiasm?

These questions require different evidence. An experiment with firms can inform a hypothesis about decision discipline without establishing child learning. A provider page can establish that a course is described and offered without establishing its effects. A completed in-game notebook can show what a student wrote without proving independent reasoning. EAG product definition establishes intended product boundaries but supplies no outcome data.

This report consequently treats “entrepreneurial skill” as a label requiring inspection. Self-reported persistence, tested knowledge, interest in founding a business and the decision to abandon an idea are different constructs. A single success score would conceal those distinctions. The relevant initial outcome for Launch is whether a learner can connect an observation to a specific assumption and qualify the conclusion. General reasoning, durable learning and commercial competence would need additional evidence.

3. Need and context: practicing uncertainty is a narrower claim than creating founders

Different settings answer different questions
Qualitative study-context comparison. Populations, programs and outcomes differ; no common effect scale or Launch outcome is implied.
These equivalents accompany the reader figures. Source IDs refer to the report references. A0 identifies EAG's product definition reviewed for the original report, not an external study or proof of availability. Other source IDs resolve to this report's reference and access record.

Figure 1. Different settings answer different questions

Qualitative study-context comparison. Populations, programs and outcomes differ; no common effect scale or Launch outcome is implied.

Primary school | S2: Self-reports and knowledge. Facilitated Dutch child program: selected short-term self-report changes, no clear knowledge improvement. Not Launch.

Vocational college | S3: Skills and intentions. Quasi-experimental program evaluation: no significant skill improvement and lower intention. Not a child-game trial.

Adult firms | S1: Stopping and revising ideas. Randomized decision training: more idea termination; qualified pivot results. Does not establish child learning.

Sources: S1, S2, S3.
S1: https://repositorio.ie.edu/server/api/core/bitstreams/f7a19247-bc6c-4639-89b8-9b94e593172e/content
S2: https://docs.iza.org/dp6512.pdf
S3: https://papers.tinbergen.nl/08038.pdf

Figure 1. Qualitative context map. The studies address different populations and outcomes; none evaluates Launch After Class. Publication years do not identify the year of every intervention. Text and data alternative (#figure-text-and-data-alternatives).

An idea can contain several assumptions at once: a person experiences a problem, existing responses are inadequate, and a proposed change would help. Asking whether the whole idea is “good” gives a learner little guidance about which assumption to investigate. EAG's proposed exercise addresses that instructional problem by isolating one question and making a before-and-after explanation possible. This is a design rationale, not a measured prevalence estimate of students' difficulties.

The retained research spans three distinct settings. A Dutch primary-school study examined a facilitated entrepreneurship program. A Dutch college study examined a different mini-company program. A multi-trial adult study examined scientific decision training in firms. Their populations, activities, measures and stakes differ. They are useful together because they prevent an overly simple story about entrepreneurship education, not because their effects can be added.

The local need remains unmeasured. This review has no interviews with students or educators, no classroom observation, and no evidence of unmet demand for another business-themed activity. A school might prefer a writing exercise, an existing course, a science investigation or no additional activity. Launch would need to show that its particular fictional setting makes a useful reasoning task easier to understand or more worthwhile to practice.

Nor should lack of interest in entrepreneurship be treated as a deficit. A student may enjoy examining evidence without wanting to run a business. Another may conclude that a proposed idea is unsuitable and stop. Both can be compatible with the educational purpose. The program's name and business setting should not turn a limited exercise into a claim about a child's future identity, ambition or economic prospects.

4. Existing alternatives and the case a new activity would have to make

Choose the scope before choosing the tool
Provider capabilities are compared by scope. The notebook is a proposed comparison condition, not a measured intervention.
Figure 2. Choose the scope before choosing the tool

Provider capabilities are compared by scope. The notebook is a proposed comparison condition, not a measured intervention.

BizWorld+ | S4: Facilitated classroom program. Lesson plans, materials and support. Provider description, not independent efficacy evidence.

JA and NFTE | S5, S6: Broader supported learning. Courses or pathways with teachers, projects and other support. Local fit, cost and comparative outcomes unverified.

Matched notebook | EAG concept: Same narrow reasoning task. Use identical fictional evidence to test whether Launch interaction adds value beyond lesson content.

Sources: A0, S4, S5, S6.
A0. EAG product definition: /projects/launch-after-class
S4: https://bizworld.org/bizworldplus/
S5: https://global.ja.org/programs/ja-entrepreneurial-mindset
S6: https://nfte.com/what-we-do/

Figure 2. Alternatives differ in instructional scope. Provider descriptions establish capabilities, not a ranking of effectiveness. The matched notebook is an EAG-proposed comparator. Text and data alternative (#figure-text-and-data-alternatives).

BizWorld+ describes a classroom entrepreneurship curriculum with lesson plans, materials and support. Its public page establishes an existing facilitated option; its promotional statements do not supply an independent effect estimate. The present product should not be assumed identical to the earlier Dutch intervention discussed below. BizWorld, S4 (https://bizworld.org/bizworldplus/)

Junior Achievement describes JA Entrepreneurial Mindset as a semester-long, teacher-led high-school course, with volunteer involvement, case and project work, and supplementary self-guided resources. That is a broader educational commitment than the proposed Launch slice. The page does not establish local delivery, cost or comparative effectiveness. Junior Achievement, S5 (https://global.ja.org/programs/ja-entrepreneurial-mindset)

NFTE describes trained teachers, experiential project learning, lean-startup methods and program pathways. It is a relevant alternative for an educator seeking a broader entrepreneurship experience. Its public description is not evidence that its components can be reduced to a short game without losing educational value. Network for Teaching Entrepreneurship, S6 (https://nfte.com/what-we-do/)

An educator-created text notebook is another important comparator. It could present the same fictional problem, test options and modeled responses, then ask for the same revision. This is a proposed comparison condition, not an evaluated program. Its value is methodological: it separates the contribution of interactive presentation from the contribution of the lesson content.

Launch therefore needs a specific reason to exist. A plausible advantage is that students can inspect an earlier assumption beside new evidence and rehearse a revision without real-world financial exposure. But a teacher may already accomplish that with a worksheet and conversation. The appropriate test is not whether students prefer a polished game to an unrelated dull task. It is whether the same content becomes clearer, more usable or more educationally useful through Launch at an acceptable total cost.

The comparison must include preparation and facilitation. A nominally self-guided activity may still require an adult to explain vocabulary, troubleshoot access and correct misconceptions. Existing programs may provide support that a narrow standalone product lacks. Current prices, procurement terms, full lesson contents, accessibility and district availability were not verified for any alternative; this report makes no purchase recommendation.

5. Evidence, counterevidence and what transfers

Child education. Rosendahl Huber, Sloof and van Praag studied a Dutch primary-school entrepreneurship program delivered in 2010-2011, mainly to children ages 11-12. The original working paper's final sample included 2,413 pupils with both questionnaires, across 118 classes. It reported positive short-term changes in several self-reported non-cognitive measures, but no clear improvement in entrepreneurship knowledge. Reassignment after initial classroom randomization and immediate follow-up qualify interpretation. The program involved several facilitated days, so its findings do not isolate a one-test fictional loop or establish durable learning. This report uses the 2012 working paper, not independently checked numerical tables from the 2014 journal version. Rosendahl Huber, Sloof and van Praag, S2 (https://docs.iza.org/dp6512.pdf)

Adult decisions. Camuffo and colleagues' 2024 replication and extension combines four randomized trials involving 759 firms. Scientific decision training increased idea termination. A simple linear effect on the number of radical pivots did not reproduce consistently; the pattern of pivoting required more qualified interpretation. These are adult business decisions with real consequences, not child-learning outcomes. The paper supplies a reason to consider stopping an idea as a potentially useful decision, but it cannot validate Launch's teaching method, fictional evidence or commercial benefits. Camuffo and colleagues, S1 (https://repositorio.ie.edu/server/api/core/bitstreams/f7a19247-bc6c-4639-89b8-9b94e593172e/content)

Counterevidence from college education. Oosterbeek, van Praag and Ijsselstein's study compared Dutch vocational-college locations using an instrumental-variable difference-in-differences design, rather than random classroom assignment. The final publication abstract reports no significant improvement in self-assessed entrepreneurial skills and a negative effect on entrepreneurial intention. The earlier full working paper concerns the 2005-2006 academic year. The setting, identification assumptions and measures restrict generalization. Lower intention could reflect a more informed assessment of fit, but this report does not establish that explanation. The result challenges an automatic-benefit narrative; it does not prove that all entrepreneurship teaching is ineffective. Oosterbeek and colleagues, S3 working paper (https://papers.tinbergen.nl/08038.pdf), final institutional record (https://research.vu.nl/en/publications/the-impact-of-entrepreneurship-education-on-entrepreneurship-skil/).

EAG's inference is that evaluation should reward the quality of a decision explanation, not a preferred attitude toward business. If the score rises whenever a student keeps building, the activity has made perseverance the answer before examining the evidence. If the score rises whenever a student abandons an idea, it has made stopping the answer instead. Neither rule teaches contingent judgment. The learner should explain why the particular evidence supports continuation, revision, uncertainty or parking.

This inference remains a product hypothesis. The studies do not establish the best wording, feedback schedule or number of practice opportunities for Launch. They also do not show that a simulated customer can substitute for contact with real people. In this product, fictional evidence permits safe practice precisely because it avoids claiming that such contact occurred. That distinction must survive the visual presentation, explanation and any saved artifact.

6. Proposed mechanism and its failure points

An explanation must connect the evidence to the decision
EAG conceptual mechanism. Every transition is proposed; all customer evidence and cost/time consequences belong to a disclosed fiction.
Figure 3. An explanation must connect the evidence to the decision

EAG conceptual mechanism. Every transition is proposed; all customer evidence and cost/time consequences belong to a disclosed fiction.

Frame and test. Name one fictional problem → State an assumption and prediction → Choose a bounded test. The options and consequences remain authored model choices.

Interpret and decide. Inspect labeled modeled evidence → Explain what changed and remains unknown → Retain, revise or park the hypothesis. No response establishes real customer demand or commercial viability.

Preserve the reasoning. Keep original and revised explanations → Recover the current test after interruption → Continue through an equivalent text notebook. Reliability and access are requirements to verify, not demonstrated learning gains.

Sources: A0.
A0. EAG product definition: /projects/launch-after-class

Figure 3. EAG conceptual mechanism, not a demonstrated causal chain. All customer responses and cost/time consequences belong to a disclosed fictional model. No step establishes real demand or market validation. Text and data alternative (#figure-text-and-data-alternatives).

The proposed mechanism begins with an explicit assumption rather than an attractive finished product. A student identifies a fictional person's problem, predicts what a bounded test might reveal, then chooses among supplied tests. The model produces disclosed feedback. The student compares that feedback with the earlier expectation and records a revised or retained hypothesis, its limits, or a decision to park it.

Each link could fail. A hypothesis prompt may invite vague statements that cannot be tested. A menu may telegraph the supposedly correct test. Feedback may contradict the hypothesis so obviously that learners merely copy a correction. A reward animation may make one response feel desirable regardless of reasoning. A realistic customer portrait may obscure the fact that no person supplied the displayed response. These are foreseeable design risks, not observed Launch defects.

The model needs more than a disclaimer at the beginning. The result should identify what was supplied by the scenario, what follows from the chosen rule and what remains unknown. A student should be able to explain that a modeled response is evidence about the exercise's fictional situation. It is not evidence about people outside the exercise. A number generated by a model does not become more authoritative because it appears on a chart.

Cost and time consequences also need boundaries. Within the fiction, a slower test might reveal something that a faster test cannot. Such a tradeoff can make the decision interesting, but it must not masquerade as a real price, timetable or economic forecast. The initial exercise should expose its assumptions clearly enough that a facilitator can discuss how a different assumption would change the conclusion. Whether that visibility improves understanding needs direct testing.

7. Design concepts for the specific learning job

The following concepts illustrate possibilities for later design, not approved scenarios or implemented functionality.

A before-and-after notebook. The learner can revisit their original hypothesis next to the supplied result and their revised explanation. The useful interaction is comparison over time. A student who changes only the wording should not automatically be treated as having reconsidered the relevant assumption. The prompt could ask which part changed, what observation prompted it and what still has not been established.

An insufficient-evidence response. One possible result might fail to resolve the question. The learner could explain why another interpretation remains plausible and choose to park the hypothesis. The purpose is to make uncertainty expressible. It should not become a hidden trick in which “insufficient evidence” is always the safest scoring strategy. The appropriateness of a conclusion depends on the particular fixture and its declared limits.

A test with a stated question. Before choosing an option, a learner could identify what that option can reveal and what it cannot. A fictional preference response, for example, would not by itself establish willingness to act. This is a conceptual distinction, not a final customer-research lesson. The exact test options require design and age-fit review so that the exercise does not require background business knowledge it never teaches.

An equivalent text route. The EAG product definition requires a readable experiment notebook carrying the same modeled result. If a scene or animated presentation is eventually used, students should still be able to read, choose, inspect feedback and revise through text. The text route should preserve the reasoning task, not merely summarize what happened in an inaccessible scene. Keyboard use, reading support and interruption recovery need practical verification in later implementation.

The first slice does not need public pitches, peer voting, sales leaderboards or real outreach to test its central idea. Those features would introduce different learning goals and social pressures. Nor does the report establish that generative customer responses would be better than authored fixtures. A fixed fixture can make the model easier to inspect and compare, while its educational usefulness remains an open question.

8. Risks, equity and the burden of participation

The largest epistemic risk is misplaced confidence. A student may learn to follow the game's rules while becoming more certain about an idea than the evidence warrants. Completion, repeated play and confident language would miss that failure. The activity should make the limit of its evidence part of the reasoning task and evaluate whether the learner understands it afterward.

The business setting can also introduce unequal familiarity. Students with family experience of business may recognize vocabulary that others encounter for the first time. Reading difficulty or unfamiliar language could be mistaken for weak reasoning. These are plausible access concerns, not findings from a Launch sample. Later observation should separate understanding of the question from the ability to decode its presentation, and allow appropriate explanation support without silently changing the assessment.

Fictional problems need careful selection. A scenario should not ask a child to disclose family hardship, identify vulnerable classmates, contact strangers or gather private customer information. It should avoid treating one household's access to money, transport or technology as the ordinary experience of every customer. Review of scenario assumptions must include who is represented, whose constraints are omitted and whether the choices imply that disadvantage is simply a failure to be entrepreneurial.

Time burden includes more than minutes inside the activity. Educators may need to introduce the model, explain the fictional boundary, assist with controls and discuss ambiguous conclusions. Operators must maintain coherent fixtures and correct misleading feedback. A short task can be expensive if every completion requires individual explanation. Later evaluation should count that work, including support for people who do not finish.

Privacy and recovery are part of the learning conditions. The first exercise does not need real contact details or public ideas. If an account is later required, EAG product definition assigns identity to Runstr with product-specific authorization; this report does not verify such implementation. Losing a saved hypothesis or advancing a test twice can invalidate the learner's comparison. Those reliability requirements are distinct from educational effectiveness and must be verified separately.

9. Evaluation and conditions for changing the proposal

Separate explanation transfer and reliability
Proposed evaluation and change criteria. No results, validated assessment scale or success thresholds are claimed.
Figure 4. Separate explanation transfer and reliability

Proposed evaluation and change criteria. No results, validated assessment scale or success thresholds are claimed.

In-task explanation: Does the conclusion fit the evidence? Revise if students repeatedly treat modeled responses as real validation or copy feedback without explaining an assumption.

Unfamiliar scenario: Does reasoning extend beyond practice? Narrow learning claims if performance does not extend to new content; delayed assessment is needed for retention.

Practical access and effort: Can the same task work for participants? Repair barriers or simplify if the game adds avoidable burden compared with a matched notebook.

Sources: A0, S1, S2, S3.
A0. EAG product definition: /projects/launch-after-class
S1: https://repositorio.ie.edu/server/api/core/bitstreams/f7a19247-bc6c-4639-89b8-9b94e593172e/content
S2: https://docs.iza.org/dp6512.pdf
S3: https://papers.tinbergen.nl/08038.pdf

Figure 4. Proposed evidence and decision criteria. These are not completed results, validated scales or numerical success thresholds. Text and data alternative (#figure-text-and-data-alternatives).

The nearest useful evaluation would ask whether students can interpret the fictional evidence and explain a justified revision. It should begin with a small formative study under separately approved research arrangements, not a claim of broad efficacy. Reading demands, facilitation, privacy and meaningful consent would need their own planning. No participants were recruited for this report.

A proposed assessment could ask students to identify the assumption being tested, distinguish supplied fiction from real observation, explain what the test leaves unresolved and justify a next decision. These are candidate dimensions, not a validated instrument. Reviewers would need examples of acceptable alternative answers and a way to avoid rewarding writing fluency alone. Agreement between scorers and the comprehensibility of prompts should be investigated before claiming a learning effect.

A later comparison with a notebook containing the same scenario would test the additional contribution of interaction. Equivalent content and support matter: extra explanation in only one condition would blur the comparison. An unfamiliar follow-up scenario could examine whether reasoning extends beyond memorizing the practiced answer. Delayed assessment would be necessary before discussing retention. Neither completion nor a favorable satisfaction response answers these questions.

The proposal should change if students repeatedly describe synthetic responses as real customer validation, if confidence grows without justified inference, or if the game performs no better than a simpler equivalent activity while requiring more support. Simplifying the presentation, revising the fixture or choosing a notebook alone would all be legitimate responses. More elaborate venture features would not automatically repair a weak reasoning task.

Access findings should also affect the decision. Persistent barriers for keyboard users, students with reading support needs or people on constrained devices would require correction before a broader claim of usefulness. A working save system and exactly-once advancement can establish reliable task behavior; they cannot establish learning. Conversely, promising classroom explanations cannot excuse losing work or exposing private information.

10. Methods, limitations, references and corrections

This focused, question-led review used the current Launch product definition and entity brief to define scope, then examined original research and first-party alternatives. Sources were retrieved on September 19, 2026. It is not a systematic review, meta-analysis, curriculum endorsement or business recommendation. Research selection favored direct evidence relevant to hypothesis-based decisions and explicitly retained findings that complicate favorable entrepreneurship narratives.

The full child-study working paper, college-study working paper and adult replication PDF were retrieved. The adult journal HTML failed in this authoring pass; an institutional copy supplied full text. The college journal findings were checked against the institutional abstract and metadata, with methods examined in the earlier working paper. Final journal numerical tables for the child study were not independently checked. Current provider pages were read as descriptions, not independent evaluations.

Dates, designs and outcomes were kept separate. No effect-size comparison is plotted because the studies do not share a common educational outcome. The report contains no Launch dataset, participant interviews, observed usage, revenue forecast or commercial validation. The strongest remaining gaps are the appropriate age and reading level, quality of the chosen fixture, learner understanding of synthetic evidence, transfer beyond the practiced task and total educator burden.

Correction checks were limited to the retrieved records and source versions. No comprehensive correction or retraction database search was conducted, and absence of a visible notice is not proof that none exists. Correction history: September 19, 2026, first report revision. Readers identifying an error should provide EAG with the disputed statement, section and a supporting source. Readers can use the EAG contact route below to report an error.

References

- S1. Camuffo and colleagues. A scientific approach to entrepreneurial decision-making: Large-scale replication and extension (https://repositorio.ie.edu/server/api/core/bitstreams/f7a19247-bc6c-4639-89b8-9b94e593172e/content). Strategic Management Journal, 2024. DOI: 10.1002/smj.3580. Full institutional PDF, especially sections 5-7.
- S2. Rosendahl Huber, Sloof and van Praag. The Effect of Early Entrepreneurship Education: Evidence from a Randomized Field Experiment (https://docs.iza.org/dp6512.pdf). IZA Discussion Paper 6512, 2012; later journal publication, 2014. Working-paper design, sample and results sections.
- S3. Oosterbeek, van Praag and Ijsselstein. The impact of entrepreneurship education on entrepreneurship skills and motivation, working paper (https://papers.tinbergen.nl/08038.pdf), 2008; final institutional publication record (https://research.vu.nl/en/publications/the-impact-of-entrepreneurship-education-on-entrepreneurship-skil/), European Economic Review, 2010, DOI: 10.1016/j.euroecorev.2009.08.002.
- S4. BizWorld. BizWorld+ (https://bizworld.org/bizworldplus/). Undated provider page, retrieved September 19, 2026.
- S5. Junior Achievement. JA Entrepreneurial Mindset (https://global.ja.org/programs/ja-entrepreneurial-mindset). Undated provider page, retrieved September 19, 2026.
- S6. Network for Teaching Entrepreneurship. What We Do (https://nfte.com/what-we-do/). Undated provider page, retrieved September 19, 2026.

Product-definition source: EAG, Launch After Class product definition, especially sections 2-6 and the first-build-slice definition, reviewed September 19, 2026; matching entity research brief. These internal sources establish intended scope only and do not verify implementation, learning or availability.

Figure text and data alternatives

Complete semantic alternatives accompany each figure in this reader and its text download. Source IDs refer to the references in this report. A0 identifies EAG's product definition reviewed for the original report, not an external study or proof of availability. Other source IDs resolve to this report's reference and access record.

Corrections and entity context

Readers identifying an error can contact EAG (/contact) with the section, disputed statement and supporting source. Launch After Class (/projects/launch-after-class) provides the public entity context, not an external citation for EAG's product claims.